一、 Definition
a time series is a set of random variables(observations){X_{t}}
t是连续or离散
二、Examples
energy consumption、telephone calls、 Internet traffic
三、General approach of prediction
∥
x
1
−
x
2
∥
p
=
(
E
∣
x
1
−
x
2
∣
p
)
1
/
p
\Vert x_{1}-x_{2}\Vert_{p} = (E\vert x_{1}-x{2}\vert ^{p} )^{1/p}
∥x1−x2∥p=(E∣x1−x2∣p)1/p
举个例子:
分布(X,Y),density p(x,y)
p
(
x
,
y
)
X
,
Y
=
y
(
x
)
=
{
x
+
y
x
∈
[
0
,
1
]
,
y
∈
[
0
,
1
]
0
o
t
h
e
r
w
i
s
e
p(x,y)_{X,Y}= y(x)=\left\{ \begin{aligned} x+y \quad x\in [0,1],y\in [0,1]\\ 0 \quad otherwise\\ \end{aligned} \right.
p(x,y)X,Y=y(x)={x+yx∈[0,1],y∈[0,1]0otherwise
E ( Y ∣ X = x ) = ∫ − inf i n f y ∗ p Y ∣ X = x ( y ) d y E(Y|X=x)=\int_{-\inf}^{inf}y*p_{Y|X=x}(y) dy E(Y∣X=x)=∫−infinfy∗pY∣X=x(y)dy
p Y ∣ X = x ( y ) = p ( x , y ) X , Y p X ( x ) p_{Y|X=x}(y)=\frac{p(x,y)_{X,Y}}{p_{X}(x)} pY∣X=x(y)=pX(x)p(x,y)X,Y
p X ( x ) = ∫ − i n f i n f p X , Y ( x , y ) d y = ∫ 0 1 ( x + y ) d y = x + 1 / 2 p_{X}(x)=\int_{-inf}^{inf}p_{X,Y}(x,y)dy=\int_{0}^{1}(x+y)dy=x+1/2 pX(x)=∫−infinfpX,Y(x,y)dy=∫01(x+y)dy=x+1/2
so, we can get, when
y
∈
[
0
,
1
]
,
p
Y
∣
X
=
x
(
y
)
≠
0
,
o
t
h
e
r
w
i
s
e
,
p
Y
∣
X
=
x
(
y
)
=
0
y\in [0,1],p_{Y|X=x}(y)\neq0,otherwise,p_{Y|X=x}(y)=0
y∈[0,1],pY∣X=x(y)=0,otherwise,pY∣X=x(y)=0
then,
E
(
Y
∣
X
=
x
)
=
∫
−
inf
i
n
f
y
∗
p
Y
∣
X
=
x
(
y
)
d
y
=
∫
0
1
y
∗
x
+
y
x
+
1
/
2
d
y
=
x
/
2
+
1
/
3
x
+
1
/
2
E(Y|X=x)=\int_{-\inf}^{inf}y*p_{Y|X=x}(y) dy=\int_{0}^{1}y*\frac{x+y}{x+1/2} dy=\frac{x/2+1/3}{x+1/2}
E(Y∣X=x)=∫−infinfy∗pY∣X=x(y)dy=∫01y∗x+1/2x+ydy=x+1/2x/2+1/3
以上其实就是均值的内在含义
但如果不知道具体分布,
or nonlinear prediction
a
+
b
x
+
c
x
2
a+bx+cx^{2}
a+bx+cx2
进一步,
如果X,Y服从高斯分布,
就是以上那个function
因为这个高斯分布可以分解成 独立同分布随机变量的和
复习中心极限定理!!!!